Replaying Life's Tape - No Miracles Required

Guest Blogger John Beatty writes...

Stephen Gould’s influential thought experiment, “replaying life’s tape,” is commonly construed as a version of “sensitivity to initial conditions,” where slight differences in initial conditions lead to substantial differences in outcome. I think Gould had something more in mind; the replay experiment includes sensitivity to initial conditions, but there’s more to it than that. And the additional part fills a void that, left unfilled, is quite mysterious.

Here’s the most famous, but somewhat ambiguous (in light of other things he wrote) version, from Wonderful Life:

I call this experiment “replaying life's tape.” You press the rewind button and, making sure you thoroughly erase everything that actually happened, go back to any time and place in the past . . . . Then let the tape run again and see if the repetition looks at all like the original. (p. 48)

Gould’s expectation was that “any replay of the tape would lead evolution down a pathway radically different from the road actually taken” (p. 51).

He strengthened this at one point, specifying that the tape is to be replayed from the same event at the same time and place to which it was rewound: “from an identical starting point” (p. 14). I’ll call this the “identical starting point,” or “identical” version.

But Gould seemed to contradict himself in favor of a weaker version when he suggested that we rewind the tape back to some point in the past, then “Alter any early event, ever so slightly and without apparent importance at the time, and evolution cascades into a radically different channel” (p. 51). Elsewhere he spoke of “alterations at the outset” (p. 19). “Alter any early event,” or alter any event “at the outset,” sounds less like replaying “from an identical starting point,” and more like replaying from a different (altered) point. The “altered starting point” version sounds a lot like – really just like – sensitivity to initial conditions. In fact it sounds so much like sensitivity to initial conditions that one wonders why Gould didn’t just call it that (if that’s really what he meant). But anyway.

Elaborating on the “altered” version, we’re supposed to go back to some point in the past, characterized by a particular state of the world – particular values of particular variables. Then, when we press the play button, the initial state is somehow different from the one we rewound to; it’s slightly altered. A miracle! Well, it’s a thought experiment after all. Or perhaps we go back to some point in the past, characterized by some state of the world, and press “pause” while we alter the values of the variables somewhat, and then press “play.” Still fishy. But again it’s just a thought experiment.

Maybe it was the “altered” version that Matt Groening had in mind in the “Time and Punishment” episode of The Simpsons, with Homer altering the values of the variables. The relevant scenes begin with Homer trying to fix his toaster and inadvertently transforming it into a time machine (“What the . . . ?!”) that takes him back to the age of dinosaurs.

 There he recalls the advice of his father:

 If you ever travel back in time, don’t step on anything. Because the tiniest change can alter the future in ways you can’t imagine.

“Fine,” Homer reflects, “As long as I stand perfectly still and don’t touch anything, I won’t destroy the future.”

But of course he doesn’t stand perfectly still. He swats an insect buzzing around him.

Stupid bug! You go squish now! [GASPS] That was just one little insignificant mosquito. That can’t change the future, right?

Oh yes it can Homer. He returns to the present to find a totalitarian world in which Ned Flanders is the dictator, carrying-out a brainwashing program of “Neducation” and lobotomizing everyone who doesn’t conform. 

Homer returns to the past to make amends for his previous alteration. “This time I’m not gonna touch a thing. Mustn’t crush. Mustn’t kill. Made it.” Just then a dinosaur startles him, he stumbles backward and steps on a walking fish. “Oh. I wish, I wish I hadn’t killed that fish.” This time he returns to the present to find it pretty great overall except, horror of horrors, no one has heard of donuts.

He keeps returning to the past, altering it in some way until the world is different, to be sure, but in a way that’s suitable to him. Namely, everyone has reptilian tongues and can lap up their food without need of fork or spoon.

Is this episode a case of (repeatedly) rewinding back to some point in the past and replaying from a slightly different, Homer-altered point? No. It rewinds back to the same point, and replays from there. Subsequently, things happen that might well not have. Okay, it’s a foregone conclusion that Homer will f**k up in some way or other, but not this way vs. that. As one replay proceeds, Homer has a chance encounter with an insect. As another replay proceeds, from the same starting point, he has a chance encounter with a walking fish. These seemingly insignificant differences make very big differences later, which is the sensitivity-to-initial conditions part. But there is more than sensitivity to initial conditions going on here. From the same starting point, small differences result by chance. Sensitivity to initial conditions does not address how the initial, slight differences come about. On the “altered” version of the replay experiment, it’s a mystery how the initial differences arise. The “identical” version, on the other hand, indicates that the initially resulting differences are matters of chance. This is conveyed by replaying from the same point to which the tape was rewound and showing that, from there, different trajectories are possible, with initially small differences compounding into larger.

Of course Gould’s favorite illustration of the replay experiment was the Frank Capra movie, Wonderful Life. It replays the history of Bedford Falls with and without George Bailey, in order to show the despondent George that his life was not worthless, far from it. Proponents of the “altered” version of the replay experiment might imagine rewinding back to some point in the history of events that led to George’s birth, then replaying not from there but rather from some altered starting point that rules out his being born. Whereas, on the “identical” version we would rewind and replay from the same point in the past when the conception of George, or his birth, was possible but when it was also possible that he wouldn’t be conceived or born. Replaying the tape would then lead to George sometimes being born, sometimes not, with very big differences in outcomes for the town.

The film doesn’t exactly favor one version of the replay experiment over the other. We don’t know for sure where the tape was rewound to and replayed from. Wherever it was rewound to, it was then fast-forwarded to the climactic Christmas Eve to show George how sadly things would have turned out by then without him. The best case for the “identical” version, where we rewind to the same point – a point at which George might or might not have been born –  is that there are many other points in the movie where George just happens to be in the right place at the right time. He might not have been, but he was in a position to see the pharmacist miss-fill a prescription that would have had a lethal effect. He might not have, but was present the day his little brother fell through the ice and George was able to save him. He might not have, but did suffer an ear infection from his exposure that day, damaging his hearing and preventing him from going off to war, leaving him home to keep the family savings and loan going and to serve the working people of Bedford Falls in that way. He might not have been in town – he was supposed to have already departed on his honeymoon – when the run on the S&L started; but he was still around and was able to stem the tide. Throughout the film there are events that take place that might not have; and had an alternative possibility occurred, the difference in outcomes would have been enormous. The history of Bedford Falls could have been rewound to most any point and and replayed, and would have had a significantly different ending many times if not every time. There’s no reason to rewind, alter, and play. Just rewind and play.

The best case for the “altered” version of Wonderful Life is that replaying the history of Bedford Falls for George’s benefit was the brain child of his guardian angel, Clarence. Which suggests supernatural intervention. Which in turn makes some sense (I guess) of an otherwise miraculous alteration of the point to which the tape was rewound. And it would perhaps also provide a providential interpretation of what would otherwise seem like chance events in George’s life that were highly consequential. I’ll return to miracles shortly.

A film that very explicitly plays, rewinds, and replays from the very same starting point is Run Lola Run. Over and over, at the same place and time, Lola takes the phone call from her boyfriend, screams, and starts running to save him. Subsequent chance happenings are the little differences that lead to very different outcomes for her and Manni.

Sorry! Back to Gould and the book Wonderful Life. One of his main objectives was to bring the replay experiment to bear on the supposedly “lottery”-like decimation of organic forms in the late Cambrian, and its consequences. Among other things, he argued that Pikaia, the first known chordate, might well have been among the unlucky losers, in which case there might be no chordates, including us. How would a proponent of the “altered” version of the replay experiment bring it to bear on the Pikaia-human thesis? Presumably by rewinding the tape back to Pikaia’s survival amid the decimation, and replaying it not from there but from Pikaia’s demise instead. That’s definitely not what Gould had in mind:

My key experiment in replaying the tape of life begins with the Burgess fauna intact and asks whether an independent act of decimation from the same starting point would yield anything like the same groups and the same history that our planet has witnessed since the Burgess maximum in organic disparity. (p. 188)

In other words, he intended that we rewind back, beyond the decimation of organic forms, somewhere closer to the earlier Cambrian “explosion” (increase) of organic forms, and replay from there. Then, via multiple replays, we would see that Pikaia might not have survived, but did, with great consequence (if you think of us as being of any great consequence). Why rewind back to some point and alter it to get Pikaia out of the picture, when simply replaying from a point prior to the decimation would often have the same effect, with no miracles required, just the luck of the draw? What is the point of altering? I have a suggestion that I’ll get to when I return, as promised, to miracles. Also, how does the “altered” version address the lottery-like decimation of forms that Gould emphasizes? It doesn’t. It says zip all about whether Pikaia’s survival was a matter of luck or not. Worse, it buries the question and tampers with the evidence. It’s the “tampered” interpretation.

A reconstruction of Pikaia

A reconstruction of Pikaia

One more thing before I get to “Why, why alter?” and miracles and the like. Toward the end of Wonderful Life, Gould lamented that the tape can never really be replayed. Okay, but there’s a decent next-best-thing, namely to conduct simultaneous plays/replays of the evolution of initially identical populations maintained in identical (or identically modified) environments. The most enterprising and influential of such efforts is the “Longterm Evolution Experiment” conducted by Richard Lenski and his students and other collaborators. The experiment involves twelve, initially identical (cloned) populations of E. coli, as they evolve in identical (and identically altered) chemostat environments. The investigators have detected a number of differences in evolutionary outcomes among the twelve lines, differences that cannot be attributed to initial genetic differences, nor to different selection pressures (since the groups have faced identical selection pressures in their identical environments), but that seem instead to depend on chance differences in the variations (and order of variations) that have arisen in the different lineages.

One interesting difference in outcomes involves an adaptive opportunity built into the experiment from the beginning. The twelve cloned populations were grown on media that included citrate, which E. coli was known not to metabolize. But the investigators considered it within the realm of possibility that the bacteria might evolve the ability to make use of citrate as a carbon source. As of 30,000 generations, none of the twelve populations had done so. But by 31,500 generations, one lineage had succeeded.

This left the question whether the one population had experienced an extremely rare mutation that would ultimately occur in the other populations as well, rendering the difference in outcomes only temporary. Or whether the population in question had by that time, through a series of happenstance events, evolved to become uniquely capable of taking the final evolutionary steps in the direction of citrate metabolism. The investigators were able to discriminate between the possibilities by employing the frozen fossil record” of evolution up to that point. That is, after every 500 generations, samples of each lineage had been frozen. So the researchers were able to back up to a point in time in the history of that lineage, prior to the evolution of citrate metabolism, thaw the ancestors, and replay its evolution multiple times from there. And what they found was that the ability to metabolize citrate arose over and over again, suggesting that, by this point, the lineage in question had become uniquely capable of making this evolutionary breakthrough.           

A nice combination of replays that diverge and replays that do not.

The authors’ literary conclusion draws from the last passage of the Origin of Species (”Even from so simple a beginning . . .”) and from Robert Frost:

Even from so simple a beginning, small happenstances of history may lead populations along different evolutionary paths. A potentiated cell took the one less traveled by, and that has made all the difference. (Zachary Blount et al., “Historical contingency and the evolution of a key innovation in an experimental population of Escherichia coli.” PNAS 105: 7899-7906, p. 7905)

Lenski’s and his collaborators’ project is supposed to exemplify a real replaying of life’s tape. That’s why the twelve plays/replays begin from identical starting points. That’s why they do not begin from slightly altered starting points.

Why alter? Why alter? Why alter? I mentioned miracles. The connection involves causation and determinism. Sensitivity of outcomes to initial conditions – which is at least part of the replay thought experiment – is in keeping with a prominent notion of causation as “counterfactual difference-making.” To say that antecedent event A1 caused outcome O1 is to say that, had A1 not occurred – had A2 occurred instead – O1 would not have resulted. The occurrence of A1 vs. A2 makes a difference. Where sensitivity to initial conditions goes further is to suggest that the occurrence of A1 vs. a slightly different A2 makes a big difference. But that’s not the important thing for now.

The important thing is to see how differently the criteria for counterfactual difference-making can be satisfied. Proponents of counterfactual notions of causation are understandably concerned to juxtapose what actually happens or happened with counterfactual situations that are relevantly similar; realistic if not real. To attribute the extinction of dinosaurs to an asteroid impact is to say, among many other things, that had the asteroid been called back at the very last second by the extraterrestrials who sent it, then the dinos would have lasted much longer. But that counterfactual is too unrealistic and is of little help when it comes to making sense of what actually happened. There’s a tradition of juxtaposing what happened with counterfactuals that not only take for granted the actual laws of nature but everything else that has happened in the real world up to and except for the putative causal event. Now, for a determinist, not only is it not the case, but it could not possibly be the case that the events of this world transpire exactly as they have, governed by our laws of nature, up to the event in question, at which point something else happens instead. But not to worry! The counterfactual world is not our world; it’s an alternative “possible world” like ours in all the above respects up to and except for the putative causal event. However, for determinists, the problem does not go away so easily, because events could no more play out in this way in the alternative deterministic world than in our deterministic world. Thus, proponents of this approach attribute to the alternative world what they dare not attribute to ours, namely “miracles” (yes) – “minor miracles” to be sure, but miracles nonetheless. There is a price to pay for determinism! Indeterminism also has its costs, to be sure, but the indeterminist has no problem with a counterfactual scenario in which events transpire in a specified way (according to stochastic laws of nature) up to a point where one of two or more alternative events could happen next. One needn’t resort to miracles happening in alternative worlds for appropriate counterfactual situations.

Doesn’t the determinist’s pickle sound like the predicament faced by proponents of the “altered” replay experiment, i.e., having to invoke miraculous or otherwise fishy alterations of the events to which the tape is rewound, so that the replay begins from a different starting point? The source of the problem may be the same. Determinism has been a major motivation for proponents of sensitivity to initial conditions. Sensitivity makes sense of the practical unpredictability of so many phenomena, but without abandoning determinism. Paraphrasing Edward Lorenz, the present determines the future, its just that the approximate present does not approximately determine the future.

Proponents of the “altered” version may think they’re doing Gould a favor by not attributing to him the kind of indeterminism that the “identical” version seems to embrace. But I don’t think he would have appreciated the generosity. Consider his last thoughts on related issues in The Structure of Evolutionary Theory. In the epilog to the final chapter he bemoaned the pervasive conception of scientific understanding that acknowledges the importance of initial conditions, together with laws of nature, but does not regard “the resolution of such details [the initial conditions] as essential or causal components of the explanation itself” (pp. 1332-1333). It’s not immediately clear what he meant by “the resolution” of the initial conditions, but I think he was saying that it matters to our understanding of the outcome how the initial conditions came about and especially whether they were matters of chance.

The line just quoted is followed by a parenthetical diatribe about how his undergraduates typically responded to the idea of real chance by parroting Laplace, insisting that the appearance of chance is just a matter of ignorance (this does seem like an undergraduate obsession), and moreover that “if science works at all, [it must] be truly deterministic” (p. 1333). To which he responded,

Natural historians have too often been apologetic, but most emphatically should not be in supporting a plurality of legitimately scientific modes, including a narrative or historical style that explicitly links the explanation of outcomes not only to spatiotemporally invariant laws of nature, but also, if not primarily, to the specific contingencies [happenstance] of antecedent states [initial conditions], which, if constituted differently, could not have generated the observed result. (p. 1333; my italics)

“The specific contingencies of the antecedent states” are not addressed by sensitivity to initial conditions. The question of their contingency is not only ignored, but the evidence is effaced by the “altered” version of the replay experiment. Strangely effaced. And at the cost of realistic counterfactual conditions for understanding what actually transpired.

There’s no need to rewind, alter in some miraculous or otherwise sketchy way, and then play. Just rewind and play. And enjoy. But watch out!



John works in history and philosophy of science, especially evolutionary biology, and also issues concerning scientific authority, and relationships between science and the state. With regard to the issues that matter most here - on this wonderful "Extinct" blog - John is especially interested and perplexed by things having to do with chance.